Washington University School of MedicineDigital Commons@Becker

Open Access Publications

2015

Quantitative autism symptom patterns recapitulatedifferential mechanisms of genetic transmission insingle and multiple incidence familiesThomas FrazierCleveland Clinic

Eric A. YoungstromUniversity of North Carolina at Chapel Hill

Antonio Y. HardanStanford University

Stelios GeorgiadesMcMaster University

John N. ConstantinoWashington University School of Medicine in St. Louis

See next page for additional authors

Follow this and additional works at: http://digitalcommons.wustl.edu/open_access_pubs

This Open Access Publication is brought to you for free and open access by Digital Commons@Becker. It has been accepted for inclusion in OpenAccess Publications by an authorized administrator of Digital Commons@Becker. For more information, please contact engeszer@wustl.edu.

Recommended CitationFrazier, Thomas; Youngstrom, Eric A.; Hardan, Antonio Y.; Georgiades, Stelios; Constantino, John N.; and Eng, Charis, ,"Quantitativeautism symptom patterns recapitulate differential mechanisms of genetic transmission in single and multiple incidence families."Molecular Autism.6,. 58. (2015).http://digitalcommons.wustl.edu/open_access_pubs/4363

AuthorsThomas Frazier, Eric A. Youngstrom, Antonio Y. Hardan, Stelios Georgiades, John N. Constantino, and CharisEng

This open access publication is available at Digital Commons@Becker: http://digitalcommons.wustl.edu/open_access_pubs/4363

Quantitative autism symptom patternsrecapitulate differential mechanisms of genetictransmission in single and multiple incidencefamilies

Frazier et al. Molecular Autism (2015) 6:58 DOI 10.1186/s13229-015-0050-z

RESEARCH Open Access

Quantitative autism symptom patternsrecapitulate differential mechanisms ofgenetic transmission in single and multipleincidence familiesThomas W. Frazier1*, Eric A. Youngstrom2, Antonio Y. Hardan3, Stelios Georgiades4, John N. Constantino5

and Charis Eng6

Abstract

Background: Previous studies have demonstrated aggregation of autistic traits in undiagnosed family membersof children with autism spectrum disorder (ASD), which has significant implications for ASD risk in their offspring.This study capitalizes upon a large, quantitatively characterized clinical-epidemiologic family sample to establish theextent to which family transmission pattern and sex modulate ASD trait aggregation.

Methods: Data were analyzed from 5515 siblings (2657 non-ASD and 2858 ASD) included in the Interactive AutismNetwork. Autism symptom levels were measured using the Social Responsiveness Scale (SRS) and by computingDiagnostic and Statistical Manual of Mental Disorders—Fifth Edition (DSM-5) symptom scores based on items fromthe SRS and Social Communication Questionnaire. Generalized estimating equation models evaluated the influenceof family incidence types (single versus multiple incidence families; male-only ASD-affected families versus familieswith female ASD-affected children), diagnostic group (non-ASD children with and without a history of languagedelay with autistic speech and ASD-affected children), and sibling sex on ASD symptom levels.

Results: Non-ASD children manifested elevated ASD symptom burden when they were members of multipleincidence families—this effect was accentuated for male children in female ASD-containing families—or when theyhad a history of language delay with autistic qualities of speech. In this sample, ASD-affected children from multipleincidence families had lower symptom levels than their counterparts in single incidence families. Recurrence risk forASD was higher for children from female ASD-containing families than for children from male-only families.

Conclusions: Sex and patterns of family transmission modulate the risk of autism symptom burden inundiagnosed siblings of ASD-affected children. Identification of these symptoms/traits and their moleculargenetic causes may have significant implications for genetic counseling and for understanding inheritedliabilities that confer risk for ASD in successive generations. Autism symptom elevations were more dramaticin non-ASD children from multiple incidence families and those with a history of language delay and autisticqualities of speech, identifying sub-groups at substantially greater transmission risk. Higher symptom burdenand greater recurrence in children from female ASD-containing families indicate that familial aggregationpatterns are further qualified by sex-specific thresholds, supportive of the notion that females require a higherburden of deleterious liability to cross into categorical ASD diagnosis.

Keywords: Autism spectrum disorder, Multiple incidence families, Genetic epidemiology, Autism symptoms,DSM-5

* Correspondence: fraziet2@ccf.org1Center for Autism (CRS10), Pediatric Institute, Cleveland Clinic, 2801 MartinLuther King Jr. Drive, Cleveland, OH 44104, USAFull list of author information is available at the end of the article

© 2015 Frazier et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Frazier et al. Molecular Autism (2015) 6:58 DOI 10.1186/s13229-015-0050-z

BackgroundAutism spectrum disorder (ASD) is a phenotypically andetiologically heterogeneous neurodevelopmental disorder[1] with common core features of social communication/interaction impairment and the presence of restricted/re-petitive behaviors [2]. Behavioral genetic studies haveidentified a strong genetic component [3, 4], derived froma wide range of genomic mechanisms [5], including pos-sible gene-environment interactions. Family-based studiesof autism symptoms have found evidence of inter-generational transmission via familial aggregation ofelevated scores. At least a subset of unaffected familymembers from multiple incidence families, most often sib-lings and fathers [6], have greater social cognitive [7, 8]and autism trait burden [9–11], and parents with subtleelevations of autism traits are at greater risk of having anASD-affected child [12].Recent genetic studies support phenotypic findings,

identifying greater additive genetic burden in multipleincidence families [13]. However, the exact pattern ofelevated symptom burden in non-ASD children fromsingle and multiple incidence families remains uncertain.This leads to several important questions. Do elevatedsymptom levels in non-ASD children from multiple inci-dence families vary across different autism symptomdomains, as would be predicted from studies implyingdifferent etiological influences on social and repetitivebehavior symptoms [14, 15]? Or is a general predispos-ition observed toward social deficits and restricted/re-petitive behavior [3]? Do ASD-affected children frommultiple incidence families also show a more severesymptom burden relative to their counterparts from sin-gle incidence families? Answers to these questions willprovide important information about autism transmissionrisk and may provide clues to differences in the mixturesof etiologic mechanisms between these incidence patterns.Furthermore, unaffected siblings who have a history oflanguage delay with autistic qualities of speech (HLDAS)have greater autism symptom burden [9]. ExaminingHLDAS in both single and multiple incidence families isimportant for determining whether increased etiologicburden in unaffected siblings extends beyond multipleincidence families to a subset of single incidence familiesand whether HLDAS further amplifies risk.Family-based studies can also be helpful for understand-

ing how sex of the affected child influences autism symp-tom levels. The best known risk factor for autism is beingmale [16], and increasingly disparate sex ratios are ob-served at the higher functioning end of the autismspectrum [17–19]. An early study focusing on categoricalASD diagnoses in families did not suggest increasedgenetic loading or sex-specific thresholds in families con-taining ASD-affected females (female ASD-containingfamilies) [20]. However, more recently, a study of male-

only and female ASD-containing families found evidenceof higher thresholds for restricted/repetitive behavior infemales [21]. Additionally, using the Autism GeneticResource Exchange cohort, Werling and Geschwind iden-tified higher recurrence rates in siblings of female pro-bands [22]. Studies of copy number variation have furthersupported the idea of sex-specific thresholds, identifying ahigher mutational burden in ASD-affected females [23]and differential genetic and hormonal factors that maycontribute to sex-specific thresholds [24]. No previousstudies have investigated whether increased phenotypicburden in female ASD-containing families extends to bothnon-ASD and ASD-affected children, if the pattern differsacross single and multiple incidence families or if autismsymptom measurements yield distinct patterns acrossspecific Diagnostic and Statistical Manual of MentalDisorders—Fifth Edition (DSM-5) symptom criteria.Using the large Interactive Autism Network (IAN) data-

base, the present study investigated these three major fac-tors (single vs. multiple incidence families, HLDAS inunaffected children, and male-only vs. female ASD-containing families) to provide a more detailed picture ofgenetic burden and transmission risk. Analyses firstfocused on autism symptom levels in non-ASD and ASD-affected children from single and multiple incidence fam-ilies. We anticipated that ASD-affected children frommultiple incidence families would have lower symptomlevels than their counterparts from single incidence fam-ilies, opposite the pattern of non-ASD children [9], imply-ing a unique etiologic mixture rather than greater overallsymptom and genetic burden. Next, we examined whethera history of language delay with autistic qualities of speechin non-ASD siblings resulted in increased symptom bur-den. We expected that non-ASD children with HLDASwould show higher symptom levels than non-ASD chil-dren without HLDAS and that HLDAS would add to theenrichment expected for multiple incidence families.Finally, we evaluated whether the sex of the ASD-affectedchild (male-only versus female ASD-containing families)further modified autism symptom levels. Non-ASD chil-dren from female ASD-containing families were predictedto show higher symptom levels than their counterpartsfrom male-only families and recurrence risk was expectedto be higher for next-born siblings in female ASD-containing families.

MethodsParticipantsData were obtained from the IAN (http://www.ianproject.org), an Internet-based registry for families withone or more ASD-affected children (Data Export IDIAN_DATA_2013-01-28). Families were eligible for enroll-ment in IAN if the parent or legal guardian who providedinformation was English speaking, the family lived in the

Frazier et al. Molecular Autism (2015) 6:58 Page 2 of 11

USA, and their child was diagnosed with ASD by a profes-sional. To be included in the present study, caregiversmust have reported data for at least one ASD-affectedchild and at least one additional child. To evaluate symp-tom levels across child and family characteristics, a subsetof the IAN registry was used with complete data on boththe Social Communication Questionnaire (SCQ) and theSocial Responsiveness Scale (SRS). Previous analyses haveshown that this subset is highly similar to the larger IANregistry with a few exceptions: the SRS/SCQ subset isolder due to SRS administration beginning at age 4, hasfewer non-verbal individuals likely as a result of severalSRS items tapping the use of language, and has agreater proportion of white/non-Hispanic youth [25].Each of these differences was found to be very small inmagnitude (r < .10).

ProceduresInformed consent was obtained from parents/guardiansfor all participants prior to entry into the IAN data col-lection. The procedures of IAN and the present studywere reviewed and approved by the institutional reviewboards of Kennedy Krieger Institute and ClevelandClinic, respectively.Demographics, clinical diagnoses, and autism symptoms

were provided by caregivers as part of participation inIAN. Demographics included: age, sex, and race/ethnicity(coded as white non-Hispanic and other race/ethnicity).Each family was coded for two characteristics: (1)family incidence type and (2) family sex type. To codefamilies, an index child was first identified by locatingthe oldest ASD-affected child associated with eachunique IAN family identifier. Single incidence familieswere coded if at least one non-ASD (unaffected) childand no other ASD-affected children beyond the indexcase were associated with the same IAN family identi-fier. Multiple incidence families were coded if at leastone additional ASD-affected child beyond the indexcase was associated with the same IAN family identi-fier. Female ASD-containing families were coded if atleast one female ASD-affected child and possiblyother male ASD-affected children were present. Male-only families were coded if all ASD-affected childrenin the family were males. Parent and other relativediagnoses were not consistently available in IAN andwere not considered in family sex-type coding.Family sex-type definitions followed previous conventions[21, 22] and, in both types of families, non-ASD femaleand male children are possible. Recurrence risk was calcu-lated separately for next-born male and female siblingsand for female ASD-containing and male-only families.These calculations used the diagnostic classification (ASDvs. non-ASD) for the oldest available sibling after theindex case in all families [22].

ASD diagnosisCaregivers reported whether each registered child wasaffected by ASD and, if affected, the specific DSM-IV-TR diagnosis obtained from a previous clinical evalu-ation. In the SRS/SCQ subset of IAN, the vast majorityof clinical ASD diagnoses (93 %) were provided by adoctoral level professional or team. A substantial propor-tion of youth (74.9 %) was diagnosed using the AutismDiagnostic Interview—Revised, the Autism DiagnosticObservation Schedule, or both. Of these, 98.7 % scored inthe ASD-affected range for one or both instruments.Recent study of youth from the IAN registry providedadditional support for the validity of clinical ASD diagno-ses [26, 27]. Importantly, randomly selected verbal youthwith a score ≥12 on the SCQ were evaluated using theAutism Diagnostic Interview—Revised and by expert clin-ician observation. All but a handful of youth (98 %) wereconfirmed to have a DSM-IV-TR clinical diagnosis ofASD. ASD was coded by collapsing all DSM-IV-TR autismspectrum diagnoses into a single category consistent withthe Center for Disease Control epidemiological catchmentprotocols. A substantial minority of the siblings registeredas not having an ASD diagnosis have caregiver-reportedpsychiatric or developmental diagnoses (ADHD 12 %, ahistory of motor delay 7 %, an anxiety disorder 6 %, mooddisorder 4 %, and intellectual disability 0.5 %). For this rea-son and to be consistent with previous publications, werefer to these siblings as non-ASD children [25]. Previousdata have suggested that a subset of unaffected childrenfrom ASD-affected families carry a specific diathesistoward elevated autism traits [9]. On this basis, non-ASDchildren were further divided into those with a history oflanguage delay and autistic qualities of speech (non-ASDwith HLDAS) and those without (non-ASD) to investigatewhether this distinction identifies children at high risk offuture genetic transmission. HLDAS was coded as positiveif any of SCQ items 3–7 were endorsed consistent with aprevious study [9].

Autism symptomsAutism symptom constructs were obtained in two ways.The first used the total and sub-scales of the SRS [28].The SRS is 65-item, ordinally scaled (1 = “not true” to 4 =“almost always true”) quantitative assessment of the sever-ity of autism symptoms. The SRS has been extensivelyvalidated and distinguishes youth with autism from otherpsychiatric conditions [11, 28]. The present study usedSRS total raw scores to examine family incidence type andfamily sex type. In addition to the total score, SRS rawsub-scale scores were computed as item averages for thefive factor analytically derived scales: emotion recognition,social avoidance, relationships, repetitive sensory motor,and insistence on sameness [29].

Frazier et al. Molecular Autism (2015) 6:58 Page 3 of 11

DSM-5 symptomsDSM-5 symptom scores were computed using itemsfrom the SCQ and SRS that were mapped to these cri-teria [25, 30] (Additional file 1). For each DSM-5 symp-tom, counts were obtained by summing SCQ and SRSitems mapped to that symptom and then dividing by thetotal possible score. This process yielded a single, com-parable metric for each of the seven DSM-5 symptoms,ranging from 0 (no endorsement) to 1 (all items en-dorsed). Internal consistency reliability was very goodfor brief scales (α = .73–.91), indicating that any differ-ences in findings across DSM-5 symptom scales are notdue to differential reliability.

Analytic planIndependent samples t tests and Pearson’s chi-squaredtest were used to compare sample characteristics acrosssingle and multiple incidence families. When non-normaldistributions were observed, non-parametric equivalentstatistics were performed and produced the same patternof results. To facilitate effect size comparison, parametricstatistics were reported.To estimate symptom level differences between single

and multiple incidence families, generalized estimatingequation (GEE) models included individual- and family-level fixed effects factors: family incidence type (single ver-sus multiple incidence), diagnostic group (ASD versusnon-ASD with and without HLDAS), sex of the child, andtheir interactions. Age was also included as a fixed effectscovariate. SRS total raw scores, factor analytically derivedSRS sub-scale scores, and DSM-5 mapped symptomscores were dependent variables in separate analyses. Ineach model, children were clustered within families. Thetwo-way family incidence type by diagnostic group inter-action tested whether autism symptom levels differedacross single and multiple incidence families. The three-way interaction adding sex of the child evaluated whetherthis effect differed in male and female children. To explorewhether autism symptom levels were further modified bymale-only families vs. female ASD-containing families, asimilar model was computed to those above except thatfamily sex type was included as a family-level fixed effectsfactor, and models were conducted only for non-ASD sib-lings. In ASD-affected children, a simpler model was com-puted with the main effects for family incidence type,family sex type, and sex of the child due to the fact thataffected females were not present in male-only families.Independent sample t tests, means, and confidence inter-vals were used to describe the pattern of results for familysex type in ASD-affected children. SRS total raw scorewas the dependent variable in each analysis.GEE models were estimated using maximum likelihood,

and fit was considered by iteratively examining alternativecovariance structures and link functions [31]. Final models

were presented based on unstructured covariance struc-ture and a linear link function, which fit comparably andyielded a highly similar pattern of results to a Poisson log-linear link function and other covariance structures (e.g.,independent, autoregressive, and exchangeable). Age andsex of the child were included in all analyses of SRS rawscores and DSM-5 symptom scores to ensure these factorsdid not confound interpretation of the effect of familycharacteristics on autism symptom levels.To further characterize the meaning of elevated symp-

tom levels in non-ASD children, follow-up analysescomputed Pearson’s correlations estimating bivariate re-lationships between autism symptoms and developmen-tal milestones: age of independent walking, age of firstmeaningful phrase speech, and presence of motor delay.Additionally, differences in developmental milestonesbetween non-ASD siblings with and without HLDASwere computed using analysis of variance for age ofwalking and age of first use of meaningful phrase speechand chi-squared analysis for motor delay.All analyses were computed in IBM SPSS version 23.

Statistical significance was determined using p < .05.False discovery rate corrections were applied when mul-tiple autism symptom measures were examined to con-trol for type 1 error inflation due to multiple testing[32]. Given the large sample size and strong power todetect small covariate effects (power > .80 for partialrab.c = .04 and larger), the pattern and magnitude of ef-fects across autism symptom measures were also consid-ered in interpretation. Single degree of freedom F andX2 statistics were converted to Cohen’s d [33]. Cohen’s dvalues ≤.20 and .20–.50 were considered small andsmall-to-medium effects, respectively [34].

ResultsSample descriptionChildren from single and multiple incidence families didnot significantly differ on age, race/ethnicity, or the pres-ence of DSM-IV-TR diagnoses of autistic disorder orpervasive developmental disorder not otherwise specified(Table 1). Interestingly, multiple incidence families had ahigher proportion of ASD-affected children who re-ceived a diagnosis of Asperger’s disorder (21.0 %) relativeto single incidence families (15.4 %).

Single versus multiple incidence familiesReliable differences in autism symptom levels were seenbetween children from single and multiple incidence fam-ilies (Table 2; smallest p = .009 for DSM-5 B2: resistanceto change) with all measures surviving false discovery ratecorrection. The only exception was DSM-5 A2: non-verbal communication where only a trend was observed(p = .051). Additional file 2 presents the full analyticresults for SRS total raw scores.

Frazier et al. Molecular Autism (2015) 6:58 Page 4 of 11

Non-ASD children without HLDAS from multiple in-cidence families had higher levels of autism symptomson the SRS relative to non-ASD children withoutHLDAS from single incidence families (Fig. 1). This pat-tern was also generally true for DSM-5 symptoms butthe size and significance of effects was variable; signifi-cant differences were only observed for DSM-5 B2: re-sistance to change and B3: restricted interests (Fig. 2).Conversely, ASD-affected children from single incidencefamilies had significantly higher autism symptoms (Fig. 1)than ASD-affected children from multiple incidencefamilies. The effect was largest for B1: repetitive motor(Fig. 2). Inspection of scores indicates that effects weresmall, but not trivial, with SRS total raw score differ-ences of 4–8 points between children from single andmultiple incidence families. As a result, frequency distri-butions of SRS total raw scores were farther apart innon-ASD and ASD-affected children from single inci-dence families and closer together in multiple incidencefamilies (Additional file 3 presents the SRS total rawscore distributions). Females from multiple incidencefamilies show a distinctly bimodal distribution, replicating

previous findings regarding sex-specific aggregation pat-terns in multiple incidence families [35].In all non-ASD children, higher autism symptom level

was significantly and positively related to age of walkingindependently, age of first meaningful phrase speech,and the presence of a motor delay (r = .06–.21, p < .002),supporting the developmental relevance of observedsymptom elevations.

Non-ASD children with HLDASNon-ASD children with HLDAS had higher levels ofautism symptoms on the SRS (Fig. 1) and higherlevels of DSM-5 symptoms (Table 2) relative to non-ASD children without HLDAS (all p values <.001).These effects were generally medium to large in mag-nitude (Cohen’s d = .25–.84). Non-ASD children withHLDAS also showed significant delays in developmen-tal milestones relative to non-ASD children withoutHLDAS, including walking independently (14.4 vs.12.0 months; F(1,2361) = 48.45, p < .001), age of firstmeaningful phrase speech (21.0 vs. 14.5 months;F(1,2414) = 85.19, p < .001), and the presence of a

Table 1 Sample characteristics of children with autism symptom questionnaire data from single and multiple incidence families in IAN

Single incidence families Multiple incidence families Χ2, t(p)

Number of families 2262 315

Female ASD-containing (N, %) 334 (14.8 %)a 125 (39.7 %)b Χ2(1) = 117.3 (p < .001)

Male-only (N, %) 1928 (85.2 %)a 190 (60.3 %)b

Number of children 4764 751

ASD (N, %) 2262 (48.8 %)a 596 (80.6 %)b Χ2(2) = 261.1 (p < .001)

Non-ASD (N, %) 2097 (45.2 %)a 121 (16.4 %)b

Non-ASD with HLDAS 280 (6.0 %)a 22 (3.0 %)b

Age (M, SD, range) 8.7 (3.8, 4–18) 8.8 (3.6, 4–18) t(5513) = −0.6 (p = .522)

Female siblings (N, %) 1602 221

ASD (N, %) 333 (20.8 %)a 141 (63.8 %)b Χ2(2) = 187.7 (p < .001)

Non-ASD with HLDAS (N, %) 118 (7.4 %) 11 (5.0 %)

Non-ASD (N, %) 1151 (71.8 %)a 69 (31.2 %)b

Race/ethnicity

White non-Hispanic 4248 (89.2 %) 684 (91.1 %) Χ2(1) = 2.5 (p = .114)

Other or unknown 516 (10.8 %) 67 (8.9 %)

SCQ total raw (M, SD) 12.4 (11.4) 17.5 (10.3) t(5513) = −11.54 (p < .001)

SRS total T score (M, SD) 64.7 (23.8) 77.1 (21.2) t(5513) = −13.47 (p < .001)

DSM-IV-TR (N, %)

Autistic disorder 950 (42.0 %) 239 (40.1 %) Χ2(2) = 10.7 (p = .005)

PDD NOS 963 (42.6 %) 232 (38.9 %)

Asperger’s disorder 349 (15.4 %)a 125 (21.0 %)b

N = 137 non-ASD siblings could not be coded into HLDAS due to the missing data regarding the presence of language delay. The numbers of children in each categoryrepresent the number of children with complete symptom data and not the total number of children in the family. The numbers for DSM-IV-TR reflect only those with aspecific DSM-IV-TR reported diagnosis. A small proportion of ASD-affected individuals did not have a specific DSM-IV-TR diagnosis reported (ASD or PDD waslisted instead of a specific diagnosis). Lowercase letters a and b represent column proportions that are significantly different (Bonferroni adjusted p < .05).Age is presented as the child’s age when the SRS was completed

Frazier et al. Molecular Autism (2015) 6:58 Page 5 of 11

motor delay (25.8 vs. 3.1 %; X2(1) = 252.18, p < .001).Non-ASD children with HLDAS from multiple inci-dence families had only slightly higher autism symp-tom levels relative to non-ASD children with HLDASfrom single incidence families, suggesting HLDAS andfamily type do not result in additive increases in riskof genetic transmission.

Male-only versus female ASD-containing familiesLumping non-ASD male and female siblings together,there was no increase in autism symptom burden infemale ASD-containing versus male-only families(X2(1) = 0.60, p = .439; Additional file 4). However,non-ASD males from multiple incidence female ASD-containing families had significantly elevated autismsymptom levels relative to other non-ASD children(≥11 SRS total raw score points higher) (X2(1) = 6.74,p = .009; Fig. 3a; Additional file 5 presents the fullanalytic results in non-ASD children). All post hoccomparisons between non-ASD males from multipleincidence female ASD-containing families were signifi-cant (p ≤ .037), with the exception of only a weaktrend for non-ASD female children from multiple in-cidence male-only families (p = .295).

ASD-affected children from female ASD-containingfamilies had significantly higher autism symptom levelsthan ASD-affected children from male-only families(Additional file 6). Inspection of means and confidenceintervals suggests that this effect was driven by higherautism symptom levels in ASD-affected males from mul-tiple incidence female ASD-containing families relativeto other ASD-affected children from multiple incidencefamilies (≥8 SRS total raw score points higher) (smallestt(453) = 2.16, p = .031; Fig. 3b). ASD-affected males andfemales from single incidence families had comparablesymptom levels.Congruent with well-documented sex rations in ASD,

recurrence risk was significantly higher for next-bornmales (22.0 %) than next-born females (8.6 %) (RR = 2.56,95 % confidence interval (CI) = 2.15–3.07). More interest-ingly, recurrence of ASD in female ASD-containingfamilies (31.6 %) was nearly three times higher than inmale-only families (11.9 %) (RR = 2.66, 95 % CI = 2.29–3.09) (Additional file 7).

DiscussionThe present study replicated a growing literature show-ing greater autism symptom burden in non-ASD family

Table 2 Autism symptom levels in non-ASD, non-ASD with HLDAS, and ASD-affected children by family incidence type

Non-ASD siblings Non-ASD siblings with HLDAS ASD siblings Family incidencetype by diagnosticgroup

Singleincidencefamilies

Multipleincidencefamilies

Singleincidencefamilies

Multipleincidencefamilies

Singleincidencefamilies

Multipleincidencefamilies

Χ2 (p)

M (SD) M (SD) M (SD) M (SD) M (SD) M (SD)

SRS

Total raw score 19.7 (19.4) 27.9 (27.3) 41.4 (31.4) 45.5 (25.6) 105.6 (29.0) 101.0 (33.1) 17.28 (<.001)

Emotion recognition .41 (.37) .56 (.52) .83 (.56) .87 (.44) 1.85 (.48) 1.79 (.55) 14.45 (.001)

Social avoidance .19 (.31) .27 (.42) .35 (.45) .45 (.51) 1.16 (.62) 1.08 (.65) 9.28 (.010)

Interpersonal relationships .23 (.39) .38 (.55) .57 (.67) .72 (.71) 1.69 (.66) 1.59 (.71) 11.26 (.004)

Repetitive sensory motor .19 (.26) .25 (.31) .45 (.45) .48 (.36) 1.35 (.60) 1.26 (.65) 18.11 (<.001)

Insistence on sameness .29 (.33) .42 (.46) .60 (.52) .66 (.45) 1.63 (.51) 1.56 (.55) 16.75 (<.001)

DSM-5 SCI

A1: socio-emotional reciprocity .08 (.12) .12 (.14) .19 (.20) .19 (.16) .61 (.21) .59 (.24) 14.20 (.001)

A2: non-verbal communication .15 (.14) .16 (.15) .27 (.19) .29 (.16) .64 (.19) .61 (.21) 5.97 (.051)

A3: relationships .07 (.14) .11 (.18) .20 (.24) .23 (.23) .68 (.24) .63 (.27) 11.80 (.003)

DSM-5 RRB

B1: repetitive motor .03 (.10) .05 (.13) .09 (.18) .08 (.18) .60 (.30) .53 (.32) 21.74 (<.001)

B2: resistance to change .13 (.17) .19 (.24) .26 (.23) .30 (.22) .65 (.24) .64 (.24) 9.43 (.009)

B3: restricted interests .06 (.14) .11 (.16) .17 (.22) .17 (.19) .66 (.25) .62 (.27) 18.96 (<.001)

B4: abnormal sensory .05 (.11) .08 (.13) .15 (.21) .12 (.16) .53 (.25) .49 (.27) 17.96 (<.001)

Higher raw and T scores indicate higher levels of autism traits. Factor analytically derived SRS sub-scales are presented as raw item averages (range 0–3; 0 = nottrue, 1 = sometimes true, 2 = often true, 3 = almost always true). T scores are not currently available for these sub-scales. Each DSM-5 symptom score representsthe proportion of endorsed items mapped to that symptom. Scale endpoints represent: 0 = no items endorsed by any individual and 1 = all items endorsedacross all individuals

Frazier et al. Molecular Autism (2015) 6:58 Page 6 of 11

members from multiple incidence families [7–9, 12,36, 37] and generated four unique findings not previ-ously described. First, non-ASD children from mul-tiple incidence families had a specific pattern ofincreased DSM-5 symptom burden. In these children,symptom increases were strongest for behaviors thatare prominent in cognitively able individuals with aut-ism (B2: resistance to change, B3: restricted interests).This pattern cannot be explained by differential reli-ability of scores and instead may indicate that B2 andB3 symptoms are more sensitive to broad ASD-likephenotypic presentations. The pattern may also implythe presence of increased burden of inherited liabil-ities in multiple incidence families, such as commonvariation in genes important for the functioning of

cortico-striatal-thalamic circuits influencing the mani-festation of repetitive/perseverative behavior. In con-trast to non-ASD children, ASD-affected childrenfrom single incidence families showed increased levelsof DSM-5 repetitive motor symptoms associated withcognitively impaired presentations of ASD. This mayindicate a disproportionate representation of severelyaffected individuals in single incidence families, pos-sibly due to strong, de novo genetic variants (largecopy number variations (CNVs), multiple deleteriousvariants) modifying basic aspects of brain develop-ment. The lack of elevated autism symptoms in themajority of non-ASD children from single incidencefamilies, particularly those without HLDAS, furthersupports the possibility that de novo events or rare

Fig. 2 Effect sizes (Cohen’s d) representing differences between single and multiple incidence families across DSM-5 symptom domains, separately forASD and non-ASD children. *p < .05; **p < .01. Positive effect sizes represent higher symptom levels in youth from single incidence families (orangebars). Negative effect sizes represent higher symptom levels in youth from multiple incidence families (light blue bars)

Fig. 1 Autism symptoms (M +/− 95 % CI) in non-ASD, non-ASD with HLDAS, and ASD-affected children from single and multiple incidence families.HLDAS History of language delay with autistic speech

Frazier et al. Molecular Autism (2015) 6:58 Page 7 of 11

recessive mutations of large effect may be driving aut-ism in a significant proportion of single incidencefamilies [11, 38].The second novel finding was that ASD-affected chil-

dren from multiple incidence families show the oppositepattern of their non-ASD sibling counterparts—decreasedsymptom burden. Thus, in multiple incidence families,non-ASD and ASD-affected children have closer levels ofautism severity, whereas a wider spread of severity is seenin single incidence families. This pattern supports thepresence of unique etiologic mixtures in single and mul-tiple incidence families, although the exact nature of thedifferences in etiologic mixture cannot be discerned fromsymptom data. A recent study of common variants identi-fied a stronger contribution of additive genetic variationin multiple incidence families [13], while studies of denovo CNVs have suggested enrichment in single incidencefamilies [39, 40]. Subtle differences in the prevalence ofcommon variants of small effect and de novo rare variantsof large effect may be leading to small, but meaningful,differences in symptom aggregation between single andmultiple incidence families. The present results highlightthe importance of accounting for family incidence type infuture etiologic investigations.Third, HLDAS appears to be a major risk factor for

elevated, but still sub-threshold, autism symptom levels[9]. Non-ASD children with HLDAS had substantiallyhigher levels of autism symptoms (Δ18–21 SRS total rawscore points) relative to non-ASD children withoutHLDAS, and elevations were present across both socialand restricted/repetitive behavior domains. Given thispattern, non-ASD children with HLDAS, including thosefrom single incidence families, may be at substantiallygreater risk of genetic transmission of autism. Prospect-ive studies are needed to confirm this observation,examine rates of fecundity in these individuals, and esti-mate the risk of ASD in their offspring.

The final novel finding has increased autism symptomburden in next-born male siblings from multiple inci-dence female ASD-containing families and greater recur-rence risk for all siblings in female ASD-containingfamilies. This finding is congruent with Szatmari andcolleagues [21] and the sex-specific threshold model.Under this model, females may require additional dis-ease burden to cross threshold [23, 41, 42], with somefemales remaining sub-threshold [43] and possibly meet-ing DSM-5 criteria for social communication disorder orother developmental psychiatry disorders rather thanautism spectrum disorder. It is important to note thatprevious studies lumping male and female non-ASD sib-lings from female ASD-containing families found a gen-eral increase in autism symptom burden. In the presentstudy, the increase in symptom burden in siblings fromfemale ASD-containing families was attenuated andnon-significant when male and female non-ASD siblingswere lumped. It is possible that previous significant find-ings using a lumped sibling group may have been drivenby a larger increase in male non-ASD siblings. Alterna-tively, the present results may represent a false negativedue to sampling variation in female non-ASD siblings.Regardless, the present findings raise the interesting pos-sibility that all children from female ASD-containingfamilies may be at higher recurrence risk, but residualsymptom burden may be most apparent in non-ASDmales from female ASD-containing families. If true,non-ASD females may be less likely than non-ASDmales to exhibit the increase in risk behaviorally becauseof the presence of female-specific protective factors (e.g.,social, hormonal, or genetic factors). Use of blinded,clinician-rated, quantitative instruments will be essentialfor clarifying the true pattern, ruling out the possibilityof rater biases, and determining the exact nature oftransmission of symptom elevations in pedigrees fromfemale ASD-containing families.

Fig. 3 Autism symptom levels (M +/− 95 % CI) in female ASD-containing and male-only families, separately for non-ASD (left panel) and ASD-affectedchildren (right panel)

Frazier et al. Molecular Autism (2015) 6:58 Page 8 of 11

Limitations and future directionsThe primary limitations of the present study were reli-ance on caregiver-reported symptoms and the use ofdata from an Internet-based registry. The IAN registry iscarefully maintained with ease of reporting for parents.However, it is possible that some caregivers did not re-port data for an unaffected sibling. Many families in theregistry contain multiple unaffected children and, ifpresent, the most likely impact of under-reporting un-affected children is to reduce power to detect differences.Caregiver-reported symptoms are susceptible to rater

biases. For example, it is possible that the rater contrastbetween ASD-affected and non-ASD children in singleincidence families is stronger and therefore leads care-givers to generate polarized ratings. However, this possi-bility seems unlikely to account for the observed patternbecause: (1) the full range of symptom levels was ob-served in both single and multiple incidence families, (2)DSM-5 symptom increases in non-ASD children wasstronger to symptom presentations seem more fre-quently in cognitively able individuals (B2: resistance tochange and B3: restricted interests) rather than being ageneral increase across all symptoms and could not beaccounted for by differential reliability, and (3) recur-rence risks for categorical ASD diagnosis were consistentwith symptom level findings. Rater contrast effects couldalso work in the opposite direction by reducing powerand leading to underestimation of symptom level in-creases in non-ASD siblings from multiple incidencefamilies. Thus, on balance, it seems unlikely that ratercontrast effects generated the overall pattern of findings.Future studies that include clinician-rated symptoms,teacher-report data [11], or objective measurements ofcognitive processes associated with the broad autismphenotype [7] will be helpful for further uncovering rela-tionships between family structure or pedigree andphenotypic presentation.Mapping of SRS and SCQ items to DSM-5 symptoms

is a proxy for actual DSM-5 symptom ratings obtainedas part of a comprehensive diagnostic evaluation. Futurestudies using DSM-5 symptom ratings will be essentialto confirm the interesting pattern of symptom elevationsin multiple incidence families. An additional limitation isthat the IAN Internet registry is not representative of allfamilies affected by ASD, due to over-inclusion of whitenon-Hispanic children and high socioeconomic statusfamilies, and does not include data from families un-affected by ASD. The presence of data from youthwhose families are not affected by ASD might assist ininterpreting the pattern of findings for non-ASD chil-dren from single and multiple incidence families. In spiteof these limitations, the national scope and large size ofthe IAN registry—as well as the simultaneous collectionof single and multiple incidence families—make it

powerful for quantifying the effects of family structureon autism while covering the full range of autism symp-tom severity.It is important to consider that the magnitude of fam-

ily incidence and family sex effects in this study tendedto be small in size. Effect sizes representing differencesbetween single and multiple incidence families may bereduced due to stoppage—a decision to curtail futurereproduction [44], without which a proportion of singleincidence families would be re-classified as multiple inci-dence families. However, pedigree transmission studiesidentifying powerful genetic effects transmitted from un-affected mothers imply similar genetic mechanisms in atleast a subset of single and multiple incidence families[45, 46]. Clearly, additional etiologic investigations areneeded to sort out these mixtures, but the present re-sults identifying relatively modest symptom differencessuggests that etiologic mechanisms may differ in propor-tion but not in kind [39].

ConclusionsWith these limitations considered, the present findingsreplicated previous data identifying elevated autismsymptoms in non-ASD children from multiple incidencefamilies and extended these findings by showing the op-posite pattern in ASD-affected children. Aggregation ofautism symptoms was also influenced by the presence ofthe HLDAS phenotype in non-ASD youth and in familieswith an ASD-affected female, both of which increasedburden. Non-ASD youth from multiple incidence familiesand youth with HLDAS appear to be at substantially ele-vated risk of autism transmission to their offspring.Finally, elevated symptom levels and recurrence risk infemale ASD-containing families support the notion ofsex-specific thresholds where females require greater gen-etic loading to meet ASD diagnostic criteria. Future large-scale studies that carefully ascertain family pedigrees andcode for additional aspects of family structure (e.g., broadphenotype in biological parents) and include blindedclinician-rated and objective measures are needed to fur-ther clarify the genetic epidemiology of autism. Thesestudies will be necessary for tailoring genetic counselingand for understanding background inherited liabilities thatconfer risk for clinical ASD in successive generations.

Additional files

Additional file 1: Social Communication Questionnaire (SCQ) andSocial Responsiveness Scale (SRS) item mapping to DSM-5 criteria.This file provides the SCQ and SRS items which map to each of the DSM-5criteria to create DSM symptom scores.

Additional file 2: Main effect and interaction tests examining SRStotal raw score by family incidence type and diagnostic status. Thisfile provides generalized estimating equation results for main effects and

Frazier et al. Molecular Autism (2015) 6:58 Page 9 of 11

interactions examining autism symptom levels across single and multipleincidence families.

Additional file 3: Frequency distributions of SRS total raw scoresin non-ASD and ASD-affected children from single and multipleincidence families, separately for female and male children. Thisfile provides frequency distributions for SRS total raw scores by familyincidence type and sex of the child.

Additional file 4: Autism symptom levels (M + 95 % CI) in non-ASDsiblings (male and female siblings combined) from femaleASD-containing and male-only families, separately for singleand multiple incidence families. This file provides SRS total rawscores (M +/−95 % CI) in all non-ASD children by family incidence typeand family sex type.

Additional file 5: Main effect and interaction tests examining SRStotal raw score by family incidence type and family sex type innon-ASD siblings. This file provides generalized estimating equationresults for autism symptom levels across family incidence type and familysex type in non-ASD children.

Additional file 6: Main effects examining SRS total raw score byfamily incidence type and family sex type in ASD-affected siblings.This file provides generalized estimating equation results for autismsymptom levels by family incidence type and family sex type inASD-affected children.

Additional file 7: Recurrence risk of ASD in next-born siblings bysex of the sibling and family sex type. This file provides recurrence riskestimates for next-born siblings separately for male and female siblingsand for male-only and female ASD-containing families.

AbbreviationsADHD: attention deficit hyperactivity disorder; ASD: autism spectrumdisorder; DSM-5: Diagnostic and Statistical Manual of Mental Disorders—FifthEdition; DSM-IV-TR: Diagnostic and Statistical Manual of MentalDisorders—Fourth Edition text revision; GEE: generalized estimatingequations; HLDAS: history of language delay with autistic qualities of speech;IAN: Interactive Autism Network; IBM SPSS: International Business MachinesStatistical Package for the Social Sciences; SCQ: Social CommunicationQuestionnaire; SRS: Social Responsiveness Scale.

Competing interestsDr. Frazier has received federal funding or research support from, acted as aconsultant to, received travel support from, and/or received a speaker’shonorarium from the Cole Family Research Fund, Simons Foundation, IngallsFoundation, Forest Laboratories, Ecoeos, IntegraGen, Kugona LLC, ShireDevelopment, Bristol-Myers Squibb, National Institutes of Health, and theBrain and Behavior Research Foundation. Dr. John Constantino receivesroyalties from Western Psychological Services for the commercial distributionof one of the metrics used in this study, the Social Responsiveness Scale.Dr. Eng received research support from IntegraGen, is an unpaid member ofthe External Scientific Advisory Boards of Ecoeos and GenomOncology,serves on the External Strategic Advisory Board of N-of-One, and serves onthe External Advisory Boards of the Center for Personalized Medicine, MissionHealth, NC, CareSource and Medical Mutual of Ohio. Drs. Youngstrom,Hardan, and Georgiades have no competing interests to disclose. No authorshave non-financial competing interests to disclose.

Authors’ contributionsTWF, EAY, and JNC designed the present study and obtained funding tosupport analyses. TWF, EAY, AYH, JNC, SG, and CE supervised datainterpretation of the study. TWF conducted data management and dataanalyses. All authors contributed to writing and revision and all authors readand approved the final manuscript.

AcknowledgementsData used in the preparation of this article were obtained from theInteractive Autism Network (IAN) Research Database at the Hugo W. MoserResearch Institute at Kennedy Krieger Inc., version 2013-01-28. For up-to-dateinformation, see http://www.ianproject.org.The authors wish to acknowledge the important contribution of theparticipants with autism, their siblings, and their families. This publication

was made possible by the Stephan and Allison Cole Family Research Fund;Case Western Reserve University/Cleveland Clinic CTSA Grant Number UL1RR024989 from the National Center for Research Resources (NCRR), acomponent of the National Institutes of Health and NIH roadmap forMedical Research; and the Developmental Synaptopathies Consortium(U54NS092090), a part of NCATS Rare Disease Clinical Research Network(RDCRN), an initiative of the Office of Rare Disease Research (ORDR). TheDevelopmental Synaptopathies Consortium is funded through thecollaboration between NCATS and the National Institute of NeurologicalDisorders and Stroke (NINDS) of the National Institutes of Health. Thecontent of this publication is solely the responsibility of the authors anddoes not necessarily represent the official views of the National Institutes ofHealth. CE is the Sondra J. and Stephen R. Hardis Chair of Cancer GenomicMedicine at the Cleveland Clinic and an American Cancer Society ClinicalResearch Professor.

Author details1Center for Autism (CRS10), Pediatric Institute, Cleveland Clinic, 2801 MartinLuther King Jr. Drive, Cleveland, OH 44104, USA. 2Department of Psychology,University of North Carolina at Chapel Hill, 235 E. Cameron Avenue, DavieHall CB #3270, Chapel Hill, NC 27599-3520, USA. 3Department of Psychiatryand Behavioral Science, Stanford University, 401 Quarry Road, Stanford, CA94305-5717, USA. 4Department of Psychiatry and Behavioural Neurosciences,Offord Centre for Child Studies, McMaster University, 555 Sanatorium Rd,Hamilton, ON L9C 2B1, Canada. 5Department of Psychiatry, WashingtonUniversity of St. Louis, 660 S. Euclid Avenue, St. Louis, MO 63110, USA.6Genomic Medicine Institute, Taussig Cancer Institute, and Stanley ShalomZielony Institute of Nursing Excellence, Cleveland Clinic, 9500 Euclid Avenue,Cleveland, OH 44195, USA.

Received: 26 February 2015 Accepted: 15 October 2015

References1. Geschwind DH. Genetics of autism spectrum disorders. Trends Cogn Sci.

2011;15:409–16.2. American Psychiatric Association. Diagnostic and statistical manual of

mental disorders. 5th ed. Arlington, VA: American Psychiatric Association;2013.

3. Frazier TW, Thompson L, Youngstrom EA, Law P, Hardan AY, Eng C, et al.A twin study of heritable and shared environmental contributions toautism. J Autism Dev Disord. 2014;44:2013–25.

4. Ronald A, Hoekstra RA. Autism spectrum disorders and autistic traits:a decade of new twin studies. Am J Med Genet B Neuropsychiatr Genet.2011;156B:255–74.

5. Abrahams BS, Geschwind DH. Advances in autism genetics: on thethreshold of a new neurobiology. Nat Rev Genet. 2008;9:341–55.

6. Schwichtenberg AJ, Young GS, Sigman M, Hutman T, Ozonoff S. Can familyaffectedness inform infant sibling outcomes of autism spectrum disorders?J Child Psychol Psychiatry. 2010;51:1021–30.

7. Losh M, Adolphs R, Poe MD, Couture S, Penn D, Baranek GT, et al.Neuropsychological profile of autism and the broad autism phenotype.Arch Gen Psychiatry. 2009;66:518–26.

8. Losh M, Piven J. Social-cognition and broad autism phenotype: identifyinggenetically meaningful phenotypes. J Child Psychol Psychiatry. 2007;48:105–12.

9. Constantino JN, Zhang Y, Frazier TW, Abbacchi AM, Law P. Siblingrecurrence and the genetic epidemiology of autism. Am J Psychiatry.2010;167:1349–56.

10. Piven J, Palmer P, Jacohbi D, Childress D, Arndt S. Broad autism phenotype:evidence from a family history study of multiple-incidence autism families.Am J Psychiatry. 1997;154:185–90.

11. Virkud YV, Todd RD, Abbacchi AM, Zhang Y, Constantino JN. Familialaggregation of quantitative autistic traits in multiplex versus simplex autism.Am J Med Genet B Neuropsychiatr Genet. 2009;150B:328–34.

12. Constantino JN, Todd RD. Intergenerational transmission of subthresholdautistic traits in the general population. Biol Psychiatry. 2005;57:655–60.

13. Klei L, Sanders SJ, Murtha MT, Hus V, Lowe JK, Willsey AJ, et al. Commongenetic variants, acting additively, are a major source of risk for autism.Mol Autism. 2012;3:9.

Frazier et al. Molecular Autism (2015) 6:58 Page 10 of 11

14. Happe F, Ronald A. The “fractionable autism triad”: a review of evidencefrom behavioural, genetic, cognitive and neural research. Neuropsychol Rev.2008;18:287–304.

15. Happe F, Ronald A, Plomin R. Time to give up on a single explanation forautism. Nat Neurosci. 2006;9:1218–20.

16. Fombonne E, Quirke S, Hagen A. Prevalence and interpretation of recenttrends in rates of pervasive developmental disorders. Mcgill J Med.2009;12:73.

17. Blumberg SJ, Bramlett MD, Kogan MD, Schieve LA, Jones JR, Lu MC.Changes in prevalence of parent-reported autism spectrum disorder inschool-aged U.S. children: 2007 to 2011–2012. In: Book changes inprevalence of parent-reported autism spectrum disorder in school-aged U.S.children: 2007 to 2011–2012 (Editor ed.^eds.), vol. 65. City: National Centerfor Health Statistics; 2013. p. 1–11.

18. Volkmar FR, Szatmari P, Sparrow SS. Sex differences in pervasivedevelopmental disorders. J Autism Dev Disord. 1993;23:579–91.

19. Bryson SE, Clark BS, Smith IM. First report of a Canadian epidemiologicalstudy of autistic syndromes. J Child Psychol Psychiatry. 1988;29:433–45.

20. Goin-Kochel RP, Abbacchi A, Constantino JN. Lack of evidence for increasedgenetic loading for autism among families of affected females: a replicationfrom family history data in two large samples. Autism. 2007;11:279–86.

21. Szatmari P, Liu XQ, Goldberg J, Zwaigenbaum L, Paterson AD,Woodbury-Smith M, et al. Sex differences in repetitive stereotypedbehaviors in autism: implications for genetic liability. Am J Med GenetB Neuropsychiatr Genet. 2012;159B:5–12.

22. Werling DM, Geschwind DH. Recurrence rates provide evidence forsex-differential, familial genetic liability for autism spectrum disordersin multiplex families and twins. Mol Autism. 2015;6:27.

23. Jacquemont S, Coe BP, Hersch M, Duyzend MH, Krumm N, Bergmann S, etal. A higher mutational burden in females supports a “female protectivemodel” in neurodevelopmental disorders. Am J Hum Genet. 2014;94:415–25.

24. Werling DM, Geschwind DH. Sex differences in autism spectrum disorders.Curr Opin Neurol. 2013;26:146–53.

25. Frazier TW, Youngstrom EA, Speer L, Embacher R, Law P, Constantino J, etal. Validation of proposed DSM-5 criteria for autism spectrum disorder.J Am Acad Child Adolesc Psychiatry. 2012;51:28–40. e23.

26. Lee H, Marvin AR, Watson T, Piggot J, Law JK, Law PA, et al. Accuracy ofphenotyping of autistic children based on Internet implemented parentreport. Am J Med Genet B Neuropsychiatr Genet. 2010;153B:1119–26.

27. Daniels AM, Rosenberg RE, Anderson C, Law JK, Marvin AR, Law PA.Verification of parent-report of child autism spectrum disorder diagnosis toa web-based autism registry. J Autism Dev Disord. 2012;42:257–65.

28. Constantino JN, Gruber CP. Social responsiveness scale: manual. LosAngeles, CA: Western Psychological Services; 2005.

29. Frazier TW, Ratliff KR, Gruber C, Zhang Y, Law PA, Constantino JN.Confirmatory factor analytic structure and measurement invariance ofquantitative autistic traits measured by the Social Responsiveness Scale-2.Autism. 2014;18:31–44.

30. Frazier TW, Youngstrom EA, Embacher R, Hardan AY, Constantino JN, Law P,et al. Demographic and clinical correlates of autism symptom domains andautism spectrum diagnosis. Autism. 2014;18:571–82.

31. Hanley JA, Negassa A, Edwardes MD, Forrester JE. Statistical analysis ofcorrelated data using generalized estimating equations: an orientation.Am J Epidemiol. 2003;157:364–75.

32. Benjamini Y, Hochberg T. On the adaptive control of the false discovery ratein multiple testing with independent statistics. J Educ Behav Stat.2000;26:60–83.

33. Rosenthal R, Rosnow RL. Essentials of behavioral research: methods anddata analysis. 2nd ed. New York: McGraw-Hill, Inc; 1991.

34. Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed.Hillsdale, NJ: Erlbaum; 1987.

35. Gockley J, Willsey AJ, Dong S, Dougherty JD, Constantino JN, Sanders SJ.The female protective effect in autism spectrum disorder is not mediatedby a single genetic locus. Mol Autism. 2015;6:25.

36. Piven J, Palmer P. Psychiatric disorder and the broad autism phenotype:evidence from a family study of multiple-incidence autism families. Am JPsychiatry. 1999;156:557–63.

37. Constantino JN, Todorov A, Hilton C, Law P, Zhang Y, Molloy E, et al. Autismrecurrence in half siblings: strong support for genetic mechanisms oftransmission in ASD. Mol Psychiatry. 2012;18:137–8.

38. Davidson J, Goin-Kochel RP, Green-Snyder LA, Hundley RJ, Warren Z, PetersSU. Expression of the broad autism phenotype in simplex autism familiesfrom the Simons Simplex Collection. J Autism Dev Disord. 2014;44:2392–9.

39. Sebat J, Lakshmi B, Malhotra D, Troge J, Lese-Martin C, Walsh T, et al. Strongassociation of de novo copy number mutations with autism. Science.2007;316:445–9.

40. Marshall CR, Noor A, Vincent JB, Lionel AC, Feuk L, Skaug J, et al. Structuralvariation of chromosomes in autism spectrum disorder. Am J Hum Genet.2008;82:477–88.

41. Robinson EB, Lichtenstein P, Anckarsater H, Happe F, Ronald A. Examiningand interpreting the female protective effect against autistic behavior. ProcNatl Acad Sci U S A. 2013;110:5258–62.

42. Jeste SS, Geschwind DH. Disentangling the heterogeneity of autismspectrum disorder through genetic findings. Nat Rev Neurol. 2014;10:74–81.

43. Dworzynski K, Ronald A, Bolton P, Happe F. How different are girls and boysabove and below the diagnostic threshold for autism spectrum disorders?J Am Acad Child Adolesc Psychiatry. 2012;51:788–97.

44. Jones MB, Szatmari P. Stoppage rules and genetic studies of autism.J Autism Dev Disord. 1988;18:31–40.

45. Zhao X, Leotta A, Kustanovich V, Lajonchere C, Geschwind DH, Law K, et al.A unified genetic theory for sporadic and inherited autism. Proc Natl AcadSci. 2007;104:12831–6.

46. Ronemus M, Iossifov I, Levy D, Wigler M. The role of de novo mutations inthe genetics of autism spectrum disorders. Nat Rev Genet. 2014;15:133–41.

Submit your next manuscript to BioMed Centraland take full advantage of:

• Convenient online submission

• Thorough peer review

• No space constraints or color figure charges

• Immediate publication on acceptance

• Inclusion in PubMed, CAS, Scopus and Google Scholar

• Research which is freely available for redistribution

Submit your manuscript at www.biomedcentral.com/submit

Frazier et al. Molecular Autism (2015) 6:58 Page 11 of 11